Process

How a testing engagement runs.

Most clients arrive with a Drupal site that already works and a test suite that doesn't. The engagement is sized to the gap — not to a headcount target, not to a fixed-fee deliverable, and not to a methodology brand-name. The shape of the work is determined by what your suite is doing on the day we read it.

Week 1

Audit.

A senior testing engineer reads your existing Playwright or Cypress suite, runs it against your environments, and produces a one-page audit. The audit names which tests are flaky, which are abandoned, which are missing, and which CI gates would benefit most from autonomous healing.

The audit is the deliverable. If you stop here, you walk away with an honest read on the state of your test surface — written by someone who maintains the tools the rest of the Drupal community uses, not by someone selling you a fix.

Relative Cost to Correct an Error chart

Bugs that escape the test suite cost an order of magnitude more to fix in production than in CI. The audit identifies which gaps in your suite are letting them through.

Week 2

Stand up the dogfood loop.

We install ATK helpers where the audit identified gaps, then wire the autonomous-healing workflow against a single CI lane on your repository. The Claude agent watches that lane during one full release cycle. Every failure it sees is either fixed in a pull request you review and merge, or routed to a human as a website-issue ticket.

Two things matter about this stage. First, we run on real failures, not seeded ones — the dogfood loop is supposed to be a representative sample of what your team would actually see week to week. Second, scope is one CI lane, not the whole pipeline. You see the workflow behave under load before broader rollout, and you can pull the plug after one cycle if it isn't earning its keep.

What changes from "we monitor your site"

Performant Labs does not run infrastructure monitoring. We don't replace your hosting team, we don't watch your synthetics, and we don't take pages. The autonomous-healing loop runs against your test suite — not against production. When tests fail at 2am, the agent diagnoses, classifies, and either opens a PR with a candidate fix or files a website-issue ticket. The release stays release-ready.

Week 3+

Take over or hand back.

After the dogfood loop has run for a release cycle, the engagement settles into one of three shapes. Two of them are equally good outcomes; the third is occasionally what is right.

01 / Hand back

Hand back, green.

Your team takes the suite back, fixed. We deliver the cleaned-up tests, the autonomous-healing workflow as a code-owned subdirectory in your repo, and a short runbook your engineers can keep operating. We don't lock you in. If your team wants to own this, that is the correct outcome and we are not going to fight you on it.

02 / Take over

Take over, ongoing.

We keep running the suite for you. The Claude agent stays wired to the CI lane it learned on, broadens to additional lanes on a schedule we agree on, and a senior testing engineer stays on the engagement as the human counterpart. Engagements are sized to the work — not to a headcount target.

03 / Embed

Embed, instead.

If the bottleneck is people rather than tooling, we send a senior testing engineer to your team for the duration of a release. They pair with your developers, own the test strategy, and ship with your release cadence. ATK and the autonomous-healing workflow come along to the extent they're useful, but the engagement is structured around the embedded engineer, not around our tools.

What we don't do.

We don't build sites. We don't run a content migration. We don't replace your Drupal team. We don't monitor your infrastructure or take pages. We don't bill against a fixed-fee build-shop SOW.

Every engagement is testing work, scoped to the gap between the suite you have and the suite your release process needs. If the gap is small, we say so and the engagement is small. If it isn't a testing gap at all, we say so and we don't take the work.

Get started

Want a one-page audit of your testing surface?

Book a testing review

A 30-minute call with a senior engineer. We'll look at your current workflow, tell you honestly which of the engagement shapes above (if any) makes sense, and leave you with a one-page writeup. No sales pitch. No obligation.

See all engagement shapes